改进稀疏表示算法在人脸识别中的应用
|
摘要
人脸识别的主要难度在于,受到光照变化、表情变化以及遮挡的影响,会使得采集的不同人的人脸图像具有相似性。为有效解决基于稀疏表示的分类算法(Sparse Representation-based Classification,SRC)在人脸训练样本不足时会导致识别率降低和稀疏表示求解效率较低的问题,提出了基于判别性低秩分解与快速稀疏表示分类(Low Rank Recovery Fast Sparse Representation-based Classification,LRR_FSRC)的人脸识别算法。利用低秩分解理论得到低秩恢复字典以及稀疏误差字典,结合低秩分解和结构不相干理论,训练出判别性低秩类字典和稀疏误差字典,并把它们结合作为测试时所用的字典;用坐标下降法来求解稀疏系数以提高了计算效率;根据重构误差实现测试样本的分类。在YALE和ORL数据库上的实验结果表明,提出的基于LRR_FSRC的人脸识别方法具有较高的识别率和计算效率。
The main difficulty in terms of face recognition lies in the occlusion as well as the changes in lightning and expressions, both of which may result in the similarity between the different face images. The classification algorithm based on sparse representation(SRC)is a classical face recognition algorithm. However, such method has the problem that the recognition rate decreases and the sparse representation has low efficiency when face training samples are insufficient. To eliminate these drawbacks, this paper proposes a face recognition algorithm based on discriminative Low Rank Recovery Fast Sparse Representation-based Classification(LRR_FSRC). Firstly, low-rank decomposition theory is used to obtain a lowrank recovery dictionary and a sparse error dictionary. This is followed by the integrated utilization of the low-rank decomposition and structural incoherence theory, which are used to train discriminative low-rank dictionary and sparse error dictionary. The discriminative low-rank dictionary and sparse error dictionary are combined as dictionary for testing. Secondly,the method of coordinate descent is used to figure out the sparse coefficient to improve the computational efficiency. Finally,according to the reconstruction error, the classification of the test sample is achieved. Experimental results on the YALE and ORL databases show that the LRR_FSRC based face recognition method proposed in this paper has higher recognition rate and faster computational efficiency.
The main difficulty in terms of face recognition lies in the occlusion as well as the changes in lightning and expressions, both of which may result in the similarity between the different face images. The classification algorithm based on sparse representation(SRC)is a classical face recognition algorithm. However, such method has the problem that the recognition rate decreases and the sparse representation has low efficiency when face training samples are insufficient. To eliminate these drawbacks, this paper proposes a face recognition algorithm based on discriminative Low Rank Recovery Fast Sparse Representation-based Classification(LRR_FSRC). Firstly, low-rank decomposition theory is used to obtain a lowrank recovery dictionary and a sparse error dictionary. This is followed by the integrated utilization of the low-rank decomposition and structural incoherence theory, which are used to train discriminative low-rank dictionary and sparse error dictionary. The discriminative low-rank dictionary and sparse error dictionary are combined as dictionary for testing. Secondly,the method of coordinate descent is used to figure out the sparse coefficient to improve the computational efficiency. Finally,according to the reconstruction error, the classification of the test sample is achieved. Experimental results on the YALE and ORL databases show that the LRR_FSRC based face recognition method proposed in this paper has higher recognition rate and faster computational efficiency.
引文
[1]王蓟翔,张扬.基于矩阵分解的压缩感知算法研究[J].通信技术,2011,44(6):138-140.
[2]Wagner A,Wright J,Ganesh A,et al.Toward a practical face recognition system:robust alignment and illumination by sparse representation[J].IEEE Transactions on Pattern Analysis&Machine Intelligence,2012,34(2):372-386.
[3]Yang J,Zhang L,Xu Y,et al.Beyond sparsity:the role of L1-optimizer in pattern classification[J].Pattern Recognition,2012,45(3):1104-1118.
[4]Kim S J,Koh K,Lustig M,et al.An interior-point method for large-scale l1-regularized least squares[J].IEEE Journal of Selected Topics in Signal Processing,2008,1(4):606-617.
[5]Deng W,Hu J,Guo J.Extended SRC:undersampled face recognition via intraclass variant dictionary[J].IEEE Transactions on Pattern Analysis&Machine Intelligence,2012,34(9):1864-1870.
[6]王学军,王文剑,曹飞龙.基于自步学习的加权稀疏表示人脸识别方法[J].计算机应用,2017,37(11):3145-3151.
[7]Cai J F,Candès E J,Shen Z.A singular value thresholding algorithm for matrix completion[J].SIAM Journal on Optimization,2008,20(4):1956-1982.
[8]Candes E J,Li X,Ma Y,et al.Robust principal component analysis?[J].Journal of the ACM,2009,58(3).
[9]杜海顺,张旭东,侯彦东,等.一种基于低秩恢复稀疏表示分类器的人脸识别方法[J].计算机科学,2014,41(4):309-313.
[10]Li C G,Guo J,Zhang H G.Local sparse representation based classification[C]//2010 20th International Conference on Pattern Recognition,2010:649-652.
[11]Huang J B,Yang M H.Fast sparse representation with prototypes[C]//2010 IEEE Computer Society Conference on Computer Vision and Pattern Recognition,2010:3618-3625.
[12]范自柱.快速稀疏表示分类的人脸识别算法[J].计算机工程与应用,2017,53(9):1-4.
[13]Lin Z,Chen M,Ma Y.The augmented Lagrange multiplier method for exact recovery of corrupted low-rank matrices[J].arXiv preprint arXiv:1009.5055,2010.
[14]施耀.基于稀疏表示的人脸识别算法研究[D].江苏苏州:苏州大学,2016.
[15]Ramirez I,Sprechmann P,Sapiro G.Classification and clustering via dictionary learning with structured incoherence and shared features[C]//2010 IEEE Computer Society Conference on Computer Vision and Pattern Recognition,2010:3501-3508.
[16]崔益峰,李开宇,胡燕,等.联合判别性低秩类字典与稀疏误差字典学习的人脸识别[J].中国图象图形学报,2017,22(9):1222-1229.
[17]Yang J,Zhang Y.Alternating direction algorithms for l1-problems in compressive sensing[J].SIAM Journal on Scientific Computing,2009,33(1):250-278.
[18]Georghiades A S,Belhumeur P N,Kriegman D J.From few to many:illumination cone models for face recognition under variable lighting and pose[J].IEEE Trans on Pattern Anal Mach Intell,2001,23(6):643-660.
[2]Wagner A,Wright J,Ganesh A,et al.Toward a practical face recognition system:robust alignment and illumination by sparse representation[J].IEEE Transactions on Pattern Analysis&Machine Intelligence,2012,34(2):372-386.
[3]Yang J,Zhang L,Xu Y,et al.Beyond sparsity:the role of L1-optimizer in pattern classification[J].Pattern Recognition,2012,45(3):1104-1118.
[4]Kim S J,Koh K,Lustig M,et al.An interior-point method for large-scale l1-regularized least squares[J].IEEE Journal of Selected Topics in Signal Processing,2008,1(4):606-617.
[5]Deng W,Hu J,Guo J.Extended SRC:undersampled face recognition via intraclass variant dictionary[J].IEEE Transactions on Pattern Analysis&Machine Intelligence,2012,34(9):1864-1870.
[6]王学军,王文剑,曹飞龙.基于自步学习的加权稀疏表示人脸识别方法[J].计算机应用,2017,37(11):3145-3151.
[7]Cai J F,Candès E J,Shen Z.A singular value thresholding algorithm for matrix completion[J].SIAM Journal on Optimization,2008,20(4):1956-1982.
[8]Candes E J,Li X,Ma Y,et al.Robust principal component analysis?[J].Journal of the ACM,2009,58(3).
[9]杜海顺,张旭东,侯彦东,等.一种基于低秩恢复稀疏表示分类器的人脸识别方法[J].计算机科学,2014,41(4):309-313.
[10]Li C G,Guo J,Zhang H G.Local sparse representation based classification[C]//2010 20th International Conference on Pattern Recognition,2010:649-652.
[11]Huang J B,Yang M H.Fast sparse representation with prototypes[C]//2010 IEEE Computer Society Conference on Computer Vision and Pattern Recognition,2010:3618-3625.
[12]范自柱.快速稀疏表示分类的人脸识别算法[J].计算机工程与应用,2017,53(9):1-4.
[13]Lin Z,Chen M,Ma Y.The augmented Lagrange multiplier method for exact recovery of corrupted low-rank matrices[J].arXiv preprint arXiv:1009.5055,2010.
[14]施耀.基于稀疏表示的人脸识别算法研究[D].江苏苏州:苏州大学,2016.
[15]Ramirez I,Sprechmann P,Sapiro G.Classification and clustering via dictionary learning with structured incoherence and shared features[C]//2010 IEEE Computer Society Conference on Computer Vision and Pattern Recognition,2010:3501-3508.
[16]崔益峰,李开宇,胡燕,等.联合判别性低秩类字典与稀疏误差字典学习的人脸识别[J].中国图象图形学报,2017,22(9):1222-1229.
[17]Yang J,Zhang Y.Alternating direction algorithms for l1-problems in compressive sensing[J].SIAM Journal on Scientific Computing,2009,33(1):250-278.
[18]Georghiades A S,Belhumeur P N,Kriegman D J.From few to many:illumination cone models for face recognition under variable lighting and pose[J].IEEE Trans on Pattern Anal Mach Intell,2001,23(6):643-660.
© 2004-2018 中国地质图书馆版权所有 京ICP备05064691号 京公网安备11010802017129号 地址：北京市海淀区学院路29号 邮编：100083 电话：办公室：(+86 10)66554848；文献借阅、咨询服务、科技查新：66554700 |